Pixel Clock
 

Hi, all,

Quick question regarding the pixel clock and resolutions. One of my co-workers has a GF4MX 440 with a DFP pixel clock of 135mhz, and a VGA pixel clock of 350mhz. I found that when I tried to hook up a 1680x1050 LCD (for reference, the Gateway FPD2185W) to his computer, his video card couldn't handle it - the error was about the pixel clock not being high enough.

On another computer, we've got a 7800GT. The LCD works perfectly fine there, pixel clock listed as 310mhz on the DFP. The Xorg.conf file was precisely the same for both computers. (HorizSync 31-70, VertRefresh 50-160).

What I can't understand is why he can take two 1600x1200 LCDs and drive them both with no problem at 60hz. Isn't 1600x1200 _greater_ than 1680x1050? Is there some sort of driver bug at work here?

Follow-up: what PCI and AGP nVidia cards can drive DVI at this resolution? We aren't adverse to spending some money to solve this. The cheaper, the better, though.

Thanks!

-Erwos

Re: Pixel Clock
 

First some small parts of theory which hopefully clarify some things.

When using a normal CRT screen you have a electron canon which needs to illuminate all pixels on the screen each refresh. The canon draws line by line and after the end of a line it needs to move back to the start which takes a certain amount of time in which the canon does 'nothing'.

A LCD screen works differently and for instance doesn't have to 'idle' when moving to a next line. Because it doesn't have to idle (and other related reasons) a LCD screen can do with a lower pixelclock than a CRT for the same resolution and refresh rate. In general CRT modelines work fine on LCD screens. You get problems when you want to use CRT modelines for resolutions like 1600x1200 and that 1680x1050 as those can require pixelclocks higher than what is supported by your card. To work around this there's a trick called 'reduced blanking' which reduces those 'move back times (blanking). This allows you to use lower pixelclocks.

The first nvidia videocards which shipped with DVI contained a tmds/lmds output with a max pixelclock of somewhere around 140MHz. (perhaps 150MHz or 130MHz, I don't know it precisely). It was designed to atleast allow 1280x1024 which at the time was the max available resolution for LCD screens. (look for good modelines at this forum as more people have this issue)

At more modern geforce cards the max pixelclock of the DVI output is somewhere between 150MHz and 165MHz (depending on the card). This should be good enough for 1600x1200 or a little higher.

Geforce7800, Quadro cards and some Geforce6 boards (atleast the 6800 card from Apple and perhaps others) contain a dual-link DVI interface which supports twice as high pixelclocks upto a little more than 300MHz. This is mainly usefull for 30" lcd screens running at 2560x1600 which can't be supported on single link.

Re: Pixel Clock
 

> Geforce7800, Quadro cards and some Geforce6 boards (atleast the 6800 card from Apple and perhaps others) contain a dual-link DVI interface which supports twice as high pixelclocks upto a little more than 300MHz. This is mainly usefull for 30" lcd screens running at 2560x1600 which can't be supported on single link.

Does it need a special cable to do this? Or will the single DVI cable that comes with a 30" LCD screen work with a Quadro FX 4000?

Re: Pixel Clock
 

You'll need a dual-link DVI cable, plugged into a dual-link DVI capable interface to power a 30" LCD (ie. Dell and Apple's 30"). Since current 30" LCD's require dual-link DVI, I'd think they would put a dual-link DVI cable in the box, so yeah it would work.

In regards to the OP, I ran my Dell 2005FPW (1680x1050) on a GeForce4 Ti4200 using DVI in windows and using VGA in linux (upgraded to PCI-E in linux before I had a chance to use DVI in linux on the Ti4200).

Re: Pixel Clock
 

That was highly educational. I take it that Xorg doesn't have proper blanking support, then?

-Erwos

Re: Pixel Clock
 

The default modeline lists shipped with X and most distributions are general modelines which work for CRT and LCD. Reducing the blanking times is only needed when you want to use a modeline which requires a higher pixelclock than supported by the card.